Re: [PATCH v3 0/3] Introduce log.showSignature config variable

From: Mehul Jain <mehul.jain2029@gmail.com>
Date: Jun 23, 2016, 05:44 UTC
Message-ID: <CA+DCAeQUJ4D4kjR6FtTibwPO8o+wv-hQNDDj2ysJaBHGTPkHgA@mail.gmail.com>
In-Reply-To: <xmqqr3bpklsd.fsf@gitster.mtv.corp.google.com>

On Thu, Jun 23, 2016 at 2:01 AM, Junio C Hamano <gitster@pobox.com> wrote:

Show 20 quoted lines

> Mehul Jain <mehul.jain2029@gmail.com> writes:
>
>> Add a new configuratation variable "log.showSignature" for git-log
>> and related commands. "log.showSignature=true" will enable user to
>> see GPG signature by default for git-log and related commands.
>>
>> Changes compared to v2:
>>       * A preparatory patch 1/3 has been introduced so that tests
>>         in patches 2/3 and 3/3 can take advantage of it.
>
> It is unclear how this change allows the remainder to "take
> advanrage" to me.  Earlier, "signed" branch was created only when
> the GPG prerequisite is met and with this change the branch is
> always created, which is the only change as far as I can see.  But
> the tests that are added in 2 and 3 are all protected with the GPG
> prerequiste.
>
> Besides, the invocation of "git commit -S" after this change is no
> longer protected by the GPG prerequisite and it may even cause the
> 'setup' step to fail on a host without GPG.

I overlooked the GPG prerequisite when I created the "setup signed branch" test in patch 1/3. I will send a patch to rectify it ones everyone agree with the idea behind this patch.

In patch 2/3 and 3/3, there are many tests which requires a branch similar to that of "signed" branch, i.e. a branch with a commit having GPG signature. So previously in v2, I created two new branches, "test_sign" and "no_sign", which are identical to that of "signed" branch. And with these branches, I wrote the tests in patch 2/3 and 3/3.

As suggested by Eric [1], rather than creating new branches, I can take advantage of "signed" branch which already exists. So, I created a new test to separate the creation of "signed" branch from existing test "log --graph --show-signature". This was done because I do not want new tests to depend on this test. If in future someone changes this test then it will affect new tests introduced in 2/3 and 3/3.

Now the new tests and existing one ("log --graph ... ") are using a single branch "signed" to do there work.

If changing an existing test is not well justified here, then I can create setup test for new tests only, without affecting the existing test.

[1]: http://thread.gmane.org/gmane.comp.version-control.git/297648

Thanks, Mehul